Python API

Load a YAML pipeline with Collectra.from_file and run it from Python:

from collectra import Collectra

pipeline = Collectra.from_file("pipeline.yaml")
pipeline.cli_run(["images"], output="results")

cli_run accepts file and directory inputs, resolving them in the same way as the command line. run accepts an explicit list of input paths:

pipeline.run(files=["images/page001.png"], output="results")

Run one task against existing results, using the task name in your pipeline:

pipeline.cli_run(
    ["results/page001.collectra"],
    task="detect_lines",
    single=True,
)

Omit single=True to continue through downstream tasks. The selected task’s input artefacts must already exist in the result directory.

To render the pipeline graph, install the Graphviz system executable and call:

pipeline.render("pipeline-graph", render=True)

See Quickstart for pipeline configuration and Artefact display for implementing custom artefact displays.